
Local proof-of-concept scanner that detects plaintext database passwords in llama-stack initialization logs, using regex pattern matching to identify exposed credentials in log files.
Local Proof-of-Concept demonstrating plaintext database password exposure in initialization logs of llama-stack versions < 0.4.0rc3.
In affected versions of llama-stack (before 0.4.0rc3), when using the pgvector vector store provider, the PostgreSQL database password is logged in plaintext during server or vector store initialization.
This information appears in:
An attacker with local access to these logs (or via misconfiguration that exposes logs) can extract valid database credentials.
CVSS: 3.2 (Low) – AV:L / AC:H / PR:N / UI:N / S:C / C:L / I:N / A:N
Fixed in: v0.4.0rc3 and later (password is masked as ****** in logs)
Important: This is a local-only vulnerability. No remote exploitation is possible without additional misconfiguration.
# Clone the repository
git clone https://github.com/YOUR-USERNAME/llama-stack-pgvector-leak-poc.git
cd llama-stack-pgvector-leak-poc
# (optional) Create virtual environment
python3 -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate